Combining watersheds and conditional random fields for image classification
Yanchai Yang, Guitao Cao · 2013
Simultaneous image segmentation and labeling are fundamental problems in computer vision. In this paper we propose a sequential method based on conditional random fields (CRF) combined with the marker-controlled watershed transform method after classification and image enhancement of artificial structures in natural images. Firstly, we use the CRF model to determine the location of interested regions. Then on the basis of the result from the CRF, we are only concentrating on labeled region by using a dual morphological reconstruction method. Lastly, the marker-controlled watershed transform method was applied to the enhanced images. Experiments show that our method has improved the accuracy of edge detection.